AI观点发散性对团体创造表现的影响及其认知神经机制
Effects of AI viewpoint divergence on group creative performance: Cognitive?neural mechanisms
周志豪 1乔煕诺 1张文宇 1童帅 1郝宁2
作者信息
- 1. 华东师范大学心理与认知科学学院
- 2. 华东师范大学心理与认知科学学院;合肥师范学院心理与认知科学学院
- 折叠
摘要
在生成式人工智能嵌入团队协作的背景下，AI观点特征如何影响团体创造表现目前仍缺乏深入探讨。本研究设计“双人-AI”团体协作场景，要求双人与AI合作完成科学与日常两类创造性任务，操纵AI输出观点的发散性，考察其对团体创意新颖性与实用性的影响，及对AI调用策略、观点采择策略、观点语义特征、双人脑间同步和单人脑功能连接的作用。结果表明，高AI观点发散性显著提升了团体创意的新颖性，但这一增益作用仅体现在日常创造性任务中；高AI观点发散性削弱了两类创造性任务中团体创意的实用性。中介分析表明，AI排除性生成策略及人类观点语义特征在AI观点发散性影响团体创造表现的过程中发挥中介作用。脑-行为整合分析显示，前额叶脑间同步与单脑功能连接可通过AI调用策略、观点采择策略与观点语义特征等变量，与团体创意新颖性和实用性产生间接关联。本研究为理解AI观点发散性影响团体创意新颖性与实用性的认知神经机制提供了实证证据，进一步拓展了团体动机性信息加工模型在人智共创情境中的应用范围。
Abstract
Generative AI is increasingly integrated into collaborative work, yet how characteristics of AI-generated viewpoints shape group creative performance remains insufficiently understood. AI viewpoint divergence refers to the extent to which AI-generated viewpoints are distributed across semantic space and differ in content. Drawing primarily on the Motivated Information Processing in Groups model, this study examined whether high versus low AI viewpoint divergence differentially affects the novelty and usefulness of group ideas in scientific and everyday creativity tasks. It also examined whether these effects are associated with cognitive processing variables and prefrontal neural indicators during humanAI co-creation.We designed a dyadAI collaborative paradigm in which two participants worked with AI to complete one scientific and one everyday creativity task. We recruited 180 participants. After excluding one dyad because one participant did not follow the task instructions, the final sample comprised 89 dyads (44 in the high-divergence condition and 45 in the low-divergence condition). We used a 2 2 mixed design, with AI viewpoint divergence as a between-dyad factor and task type as a within-dyad factor. Group creative performance was assessed in terms of novelty and usefulness. We also measured AI utilization strategies, perspective-taking strategies, and semantic features of human-generated and group-generated ideas, and used fNIRS hyperscanning to measure prefrontal inter-brain synchronization and intra-brain functional connectivity.Manipulation checks showed that AI-generated viewpoints in the high-divergence condition were more semantically dispersed than those in the low-divergence condition. High AI viewpoint divergence increased the novelty of group ideas, but this benefit emerged primarily in the everyday creativity task. In contrast, high AI viewpoint divergence reduced the usefulness of group ideas across both task types. Mediation analyses showed that AI exclusion-based generation strategies and semantic features of human ideas mediated the association between AI viewpoint divergence and group creative performance. Brainbehavior integration analyses further indicated that prefrontal inter-brain synchronization and intra-brain functional connectivity were indirectly associated with novelty and usefulness through AI utilization strategies, perspective-taking strategies, and semantic features.These findings indicate that AI viewpoint divergence does not uniformly enhance group creativity. Instead, its effects depend on task constraints and on how groups search for, select, adopt, and semantically reorganize AI-generated information. The study extends the application of the Motivated Information Processing in Groups model to humanAI co-creation and provides empirical evidence for cognitive processing pathways and neural associations linking AI viewpoint divergence to the distinct outcomes of novelty and usefulness in group creative performance. Practically, AI support for creative collaboration should be calibrated to task openness, domain constraints, and the stage-specific demands of idea generation and evaluation.关键词
AI观点发散性/团体创造表现/人智共创/fNIRS超扫描/认知神经机制Key words
AI viewpoint divergence/group creativity/human–AI co-creation/fNIRS hyperscanning/cognitive-neural mechanisms引用本文复制引用
周志豪,乔煕诺,张文宇,童帅,郝宁.AI观点发散性对团体创造表现的影响及其认知神经机制[EB/OL].(2026-07-19)[2026-07-23].https://chinaxiv.org/abs/202607.00133.学科分类
计算技术、计算机技术
